An Evolutionary Multitasking Ant Colony Optimization Method Based on Population Diversity Control for Multimodal Transport Problems

Multimodal transport is a challenging, NP-hard problem in combinational optimization and has been solved using evolutionary algorithms, which excel at solving large-scale problems. However, few studies have used evolutionary algorithms, particularly swarm intelligence algorithms, to concurrently handle multiple multimodal transport instances. Ant colony optimization (ACO), which is a population intelligence technique that is adept at identifying the optimal paths in graphs, has been primarily used to address tasks separately rather than concurrently. Therefore, in this study, we introduce a multipopulation-based multitask environment where task-specific populations run in parallel, and ACO serves as the optimizer for each task. A variance-based population diversity measure is then calculated to characterize the distribution differences among individuals. If the population diversity of a specific task falls below a predetermined threshold, the valuable routing traits extracted from other tasks are transferred to the stagnant population. Our method is called population diversity-controlled multitask ACO (PDMTACO). We use multiple benchmark traveling salesman problem (TSP) instances at different scales to validate the efficacy of PDMTACO. Subsequently, we extend PDMTACO to address a series of multimodal transport problems. Our experimental results demonstrate that the use of information transferred by our method significantly reduces its logistics costs and carbon emissions in all multimodal transport tasks.